Clinical Heterogeneity in Orthostatic Movement Disorders: Insights from Latent Class Analysis and Disease Burden...

Orthostatic tremor (OT) and orthostatic myoclonus (OM) are rare hyperkinetic disorders characterized by unsteadiness during stance. Substantial clinical overlap limits the prognostic and therapeutic value of categorical diagnostic labels.

To assess whether a multidomain disease burden framework provides more meaningful stratification than traditional OT/OM diagnoses and to identify phenotypic subgroups.

We conducted a cross-sectional analysis of 58 patients with OT/OM who underwent multidomain assessment. A composite Disease Burden Score (DBS) was derived from five domains: symptom severity, functional limitation, comorbidity burden, medication exposure, and fall risk. Patients were classified into high- and low-burden groups. Predictors of high disease burden were evaluated using Firth logistic regression and random forest classifiers. Latent class analysis (LCA) identified subgroups, and concordance between DBS strata and latent classes was assessed. Exploratory k-means and hierarchical clustering were performed for validation.

DBS stratification distinguished high- and low-burden patients with strong accuracy (AUC = 0.96). All five domains contributed to burden classification, although individual regression coefficients were imprecise. LCA identified four subgroups: low-burden, functionally impaired, comorbidity-dominant, and globally burdened. These subgroups did not align with OT/OM diagnostic categories. Concordance between LCA classes and DBS strata was weak (Cramer's V = 0.176). Demographic variables and SF-36 quality-of-life domains did not differ across latent classes.

Multidomain clinical data enable disease-burden stratification and reveal phenotypic heterogeneity in orthostatic movement disorders. Limited correspondence with traditional diagnoses supports a spectrum-based model. DBS and LCA offer complementary frameworks for individualized assessment, warranting validation in larger and longitudinal cohorts.